Structure and Information Dynamics of Pro- and Anti-Vaccine Interaction Networks in Twitter

The spread of information and misinformation regarding the Covid-19 vaccines has greatly affected immunization efforts during the current pandemic. A rigorous quantitative analysis of online social network data can reveal the structures of pro- and anti-vaccine communities and how they propagate their messages, which can help develop strategies that promote vaccination.

In this work, we present a study of all vaccine-related conversations between Twitter users in Chile during a six-month period that includes most of the mass vaccination efforts in that country. By training a machine learning model to classify tweets in a pro- to anti-vaccine spectrum, we characterize almost 60,000 accounts, distinguishing four opinion communities (pro-vaccine, promoters, inhibitors, and anti-vaccine), to analyze the structure and dynamics of their conversation networks.

Our study shows that each community has a distinct online behavior and network topology within its members and with other communities. We also find that pro-vaccine accounts tend to communicate with likeminded accounts, whereas anti-vaccine accounts mainly target conversations with pro-vaccine accounts. These results provide lessons for the design of effective communication strategies that maximize vaccination rates.